3-195 - Leveraging Machine Learning Methods for Research on Child Development in the Big Data Era

Session Type: Paper Symposium

Integrative Statement

With booming big data, the utility of machine learning (ML) methods is gaining attention in developmental science (Rosenberg et al., 2018). Facilitated by widespread access to technological advances in computational speed, ML methods are now being used in developmental research to process raw big data (Gilmore et al., 2016), predict unobserved developmental outcomes (Whelan et al., 2014), select important features from large-scale data for optimal study design (Brick et al., 2017), and discover developmental patterns (Brandmaier et al., 2017). This symposium showcases how ML methods are pushing and being pushed by research on child development. Paper 1 provides an overview of how ML is being used in both data acquisition and analysis with examples including (1) modern data management pipelines in a study of children’s self-regulation for handling multimodal data structure, (2) analysis of rapid smartphone screenshots in a sample of low-income, Mexican-American teens, and (3) selection of predictive features for sexual maturity using longitudinal data on youth development. Paper 2 utilizes a ML classification method to predict young adult educational attainment with family experiences in adolescence using nationally representative longitudinal data. Paper 3 introduces a new method based on ML algorithms that captures divergent developmental trajectories with a newly developed R package and illustrates the method using longitudinal early childhood reading data. Together, the papers highlight the potential value of ML methods across various research activities (i.e., data acquisition and analysis and method development) and on a wide range of data types (e.g., multimodal data, screenshots, longitudinal survey data).

Sub Unit

• Methods, History, Theory

Chair

• Xiaoran Sun, The Pennsylvania State University

Discussant

• Keith F Widaman, University of California, Riverside

Individual Presentations

• Redesigning Research for a Modern Machine Learning World: Merging Data Science and Developmental Science - Presenting Author: Nilam Ram, Penn State University; Non-Presenting Author: Thomas Robinson, Stanford University; Non-Presenting Author: Byron Reeves, Stanford University

• Family Experiences in Adolescence Predict Young Adult Educational Attainment: A Machine Learning Approach - Presenting Author: Xiaoran Sun, The Pennsylvania State University; Non-Presenting Author: Nilam Ram, Penn State University

• Explorations of Individual Change Processes and Their Determinants: A Novel Approach and Remaining Challenges  - Presenting Author: Kevin Grimm, Arizona State University; Non-Presenting Author: Ross Jacobucci, University of Notre Dame; Non-Presenting Author: Gabriela Stegmann, Arizona State University; Non-Presenting Author: Sarfaraz Serang, Utah State University